A Probabilistic Agent-Based Simulation for Community Level Language Change in Different Scenarios

Abstract

We built an agent-based model (ABM) to simulate historical language change, and tested it by means of a case study on word order change in English. Our modeling approach assumes that complex patterns in population-level language change can be understood in terms of many small changes, resulting from interactions between individual agents of different populations. Each agent has a language model that changes due to contact with other agents from the same or the other population. As a result, micro-level changes (i.e. at the level of individual agents) lead to macrolevel changes (i.e. at the level of the population). We implemented, manipulated and explored the effect of learning rate, likelihood of interaction between agents from different populations, location-bound dialects, and degree of agent variation within a population. Although parts of the model leave room for fine-tuning, and external factors have yet to be combined into one model, the simulation results show that ABM is a useful tool for gaining more insight into historical language change. The ABM approach has potential for modeling word order change in English, as well as language change in general.